Table Rock Lake Water-Clarity Assessment Using Landsat Thematic Mapper Satellite Data

Scientific Investigations Report 2009–5162

U.S. Department of the Interior U.S. Geological Survey Cover illustration. Landsat satellite image from the Table Rock Lake area, southwestern Missouri and northwestern Arkansas. Table Rock Lake Water-Clarity Assessment Using Landsat Thematic Mapper Satellite Data

By Gary W. Krizanich and Michael P. Finn

Scientific Investigations Report 2009–5162

U.S. Department of the Interior U.S. Geological Survey U.S. Department of the Interior KEN SALAZAR, Secretary U.S. Geological Survey Suzette M. Kimball, Acting Director

U.S. Geological Survey, Reston, Virginia: 2009

For more information on the USGS—the Federal source for science about the Earth, its natural and living resources, natural hazards, and the environment, visit http://www.usgs.gov or call 1-888-ASK-USGS For an overview of USGS information products, including maps, imagery, and publications, visit http://www.usgs.gov/pubprod To order this and other USGS information products, visit http://store.usgs.gov

Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government.

Although this report is in the public domain, permission must be secured from the individual copyright owners to reproduce any copyrighted materials contained within this report.

Suggested citation: Krizanich, G. W., and Finn, M. P., 2009, Table Rock Lake water-clarity assessment using landsat thematic mapper satellite data: U.S. Geological Survey Scientific Investigations Report 2009-5162, 9 p. iii

Contents

Abstract...... 1 Introduction...... 1 Purpose and Scope...... 1 Background...... 1 Study Area...... 3 Methods...... 3 Results ...... 6 Summary...... 8 Acknowledgments...... 8 References Cited...... 8

Figures

1–2. Maps showing: 1. Study area location...... 2 2. Table Rock Lake sample locations...... 3 3–4. Graphs showing: 3. Relation between predicted and measured values of secchi depth in meters for Table Rock Lake...... 5 4. Distribution of field samples with respect to number of days within satellite overpass...... 6 5–6. Maps showing: 5. Model predicted trophic state for Table Rock Lake – August 2003...... 7 6. Temporal comparison of model results for the James River arm of Table Rock Lake...... 7

Tables

1. Mean Landsat brightness values for Table Rock Lake sample sites 2003 and 2005...... 4 iv

Conversion Factors, Abbreviations, and Datums

Inch/Pound to SI

Multiply By To obtain Length mile (mi) 1.609 kilometer (km) Area acre 4,047 square meter (m2) acre 0.4047 hectare (ha) acre 0.4047 square hectometer (hm2) acre 0.004047 square kilometer (km2)

SI to Inch/Pound

Multiply By To obtain Length meter (m) 3.281 foot (ft) Area square meter (m2) 0.0002471 acre hectare (ha) 2.471 acre square hectometer (hm2) 2.471 acre square kilometer (km2) 247.1 acre Vertical coordinate information is referenced to the North American Vertical Datum of 1988 (NAVD 88). Horizontal coordinate information is referenced to the North American Datum of 1983 (NAD 83). Altitude, as used in this report, refers to distance above the vertical datum. Table Rock Lake Water-Clarity Assessment Using Landsat Thematic Mapper Satellite Data

By Gary W. Krizanich and Michael P. Finn

Abstract for predicting water quality using satellite reflectance data and water clarity measured by secchi disk transparency. Water quality of Table Rock Lake in southwestern Mis- souri is assessed using Landsat Thematic Mapper satellite Purpose and Scope data. A pilot study uses multidate satellite image scenes in conjunction with physical measurements of secchi disk trans- The purpose of this report is to describe the methods and parency collected by the Lakes of Missouri Volunteer Program techniques used to develop a predictive model for estimating to construct a regression model used to estimate water clarity. Table Rock Lake water quality based on satellite observations The natural log of secchi disk transparency is the dependent combined with field-data collection. The techniques described variable in the regression and the independent variables are can be used to calculate estimates of water clarity from secchi Thematic Mapper band 1 (blue) reflectance and a ratio of the disk transparency or a lake trophic state index derived from band 1 and band 3 (red) reflectance. The regression model can these measurements. Model outputs can be applied to the be used to reliably predict water clarity anywhere within the entire water body enabling water-resource managers to iden- lake. A pixel-level lake map of predicted water clarity or com- tify areas of potential concern. puted trophic state can be produced from the model output. Information derived from this model can be used by water- resource managers to assess water quality and evaluate effects Background of changes in the watershed on water quality. Eutrophication is the process of organic enrichment of lake water with respect to time (Wetzel, 2001). It is a natural aging process experienced by all lakes. Nutrient supplies primarily are derived from atmospheric and terrestrial sources, Introduction the latter being more significant with runoff being the principal source of nutrients in lakes. Increased levels of nutrients result Table Rock Lake is a primary water resource located in increased biological productivity, which can have harmful in southwestern Missouri with tourism related activities effects to water supplies. within the watershed accounting for more than a billion dol- Trophic state is a measure of a lake’s productivity and lars in revenue each year (Missouri Department of Natural is commonly used to assess water quality. Eutrophy is a state Resources, 2005). Decreasing water clarity in Table Rock of high nutrient enrichment usually associated with declin- Lake has caused concern with water-resource managers and ing water quality. The Carlson trophic state index (1977) is a private watershed groups. The reservoir was listed in 2002, numeric index commonly used for trophic state assessment of in accordance with section 303(d) of the Federal Clean Water lakes, and is based on the relation between algal biomass and Act, as having impaired water quality because of nutrient secchi disk transparency depth. The index ranges from 0 to enrichment. Long-term monitoring of the reservoir by Univer- 100 with each increase of 10 units on the trophic state index sity of Missouri researchers documents an increase in nutrients equivalent to a doubling of algal biomass and can be computed and chlorophyll with a corresponding decrease in water clarity from chlorophyll a concentration, total phosphorus concentra- over the past 20 years (Jones and Perkins, 1999). Increases tion, or secchi disk transparency values. in nitrogen and phosphorus inputs to the reservoir have been Remote sensing methods can be effective for water- attributed to point and nonpoint sources, with major contribu- quality monitoring and water-resource assessments (Ritchie tions from wastewater treatment, urban storm water, and agri- and others, 2003). Advantages include multispectral band cultural runoff (Missouri Department of Natural Resources, capabilities, repetitive coverage, a historical archive dating 2005). This report evaluates a remote sensing methodology from the early 1970s, and the ability to cover large geographic areas with a single image scene. Early studies by Boland Table Rock Lake Water-Clarity Assessment Using Landsat Thematic Mapper Satellite Data

2 Table Rock Lake Water-Clarity Assessment Using Landsat Thematic Mapper Satellite Data

95° 90°

40°

Missouri

Table Rock Lake

35° Arkansas

Base from Environmental Systems Research Institute digital data, 1983 0 25 50 75 100 MILES Universal Transverse Mercator projection, zone 15 Horizontal coordinate information is referenced to the North American Datum of 1983 (NAD 83) 0 25 50 75 100 KILOMETERS

Figure 1. Study area location.

(1976) and Wezernak and others (1976) utilized complex from satellite observations (Lathrop and Lillesand, 1986). The multivariate statistical techniques for computation of a trophic empirical relation between secchi disk depth measurements state index based on reflectance data obtained from Landsat and Landsat TM reflectance has been used to predict region- Multispectral Scanner (MSS) data. Their studies found that wide lake water clarity (Chipman and others, 2004; Fuller and secchi disk transparency and chlorophyll a concentration could others, 2004). Previous studies of Ozark lakes have shown that be used in conjunction with Landsat mean reflectance values Landsat MSS and TM data can be effective in predicting water to accurately predict lake trophic state. Increased spatial and quality based on regression models developed from chloro- radiometric resolution of the Thematic Mapper (TM) sensor phyll a and secchi disk depth data (Krizanich, 1986; Allee and provided improved accuracy of water-quality models derived Johnson, 1999). Methods 3

Study Area Methods

Table Rock Lake, located in southwestern Missouri and Two Landsat 5 Thematic Mapper scenes (8/22/2003, northwestern Arkansas (fig. 1), is a U.S. Army Corps of Engi- 8/11/2005) were purchased from the Earth Resources Observa- neers reservoir completed in 1959 to provide flood control tion and Science (EROS) center in Sioux Falls, South Dakota. and power generation in the Upper White River Basin. The Landsat images were formatted as GeoTIFF and projected reservoir has a surface area of 43,100 acres with 745 miles of to Universal Transverse Mercator, North American Datum shoreline. It is one of four reservoirs impounding the White 1983, Zone 15. Both scenes are from Worldwide Reference River with Beaver Lake immediately upstream and Lake System 2 path 25, row 35 and were selected based on mini- Taneycomo and Bull Shoals Lake downstream. Table Rock mal scene cloud cover and closeness of the scene date to field Lake is located within the Springfield Plateau of the Ozark data collection. The scenes were processed by EROS to level Dome, an area dominated by relatively flat-lying carbonate 1P specifications, which include radiometric and geometric rocks. Land cover of the area is predominantly forest and pas- correction as well as the use of 25 ground-control points to ture, but increasing urban development within the watershed improve geometric accuracy. has the potential to affect water quality (Thorpe and others, 2005).

93°40' 93°30' 93°20'

TR13

TR12

TR12 Data collection site and identifier used in table 1 TR11

36°40' TR5

TR4.5

TR15 TR3 TR2

TR17 TR1 TR14

TR18 TR16 TR9

TR6.5 TR10

TR7 TR8 36°30'

U.S. Geological Survey 1:24,000 National Hydrography Dataset, 1999 0 5 10 MILES Universal Transverse Mercator projection, zone 15 Horizontal coordinate information is referenced to the North American Datum of 1983 (NAD 83) 0 5 10 KILOMETERS

Figure 2. Table Rock Lake sample locations. Red dots indicate Lakes of Missouri Volunteer Program data collection sites. 4 Table Rock Lake Water-Clarity Assessment Using Landsat Thematic Mapper Satellite Data

Table 1. Mean Landsat brightness values for Table Rock Lake sample sites 2003 and 2005.

[TM1 and TM3, Thematic Mapper bands] Data collection Year Pixel TM1 TM3 TM1/TM3 Secchi ln Secchi site (fig. 2) count (meters) (meters) TR1 2003 9 80 24 3.3 4.9 1.6 TR2 2003 9 80 23 3.4 4.0 1.4 TR3 2003 9 78 22 3.5 3.9 1.4 TR4.5 2003 9 78 23 3.4 2.7 1.0 TR5 2003 9 79 24 3.3 1.9 0.6 TR6.5 2003 9 82 27 3.1 1.5 0.4 TR8 2003 9 75 21 3.6 3.5 1.3 TR9 2003 9 82 27 3.1 2.4 0.9 TR10 2003 9 83 26 3.2 2.4 0.9 TR11 2003 9 79 26 3.1 1.1 0.1 TR12 2003 9 80 28 2.9 0.8 -0.3 TR13 2003 9 80 27 3.0 0.9 -0.1 TR14 2003 9 77 22 3.5 3.8 1.3 TR15 2003 9 78 24 3.3 3.2 1.2 TR16 2003 9 80 28 2.9 1.4 0.3 TR17 2003 9 79 23 3.4 3.0 1.1 TR18 2003 9 86 28 3.1 3.0 1.1

TR1 2005 9 66 17 3.8 4.1 1.4 TR2 2005 9 67 17 3.8 2.0 0.7 TR3 2005 9 65 16 4.0 3.6 1.3 TR4.5 2005 9 65 16 4.0 2.6 1.0 TR5 2005 9 64 18 3.6 1.3 0.3 TR6.5 2005 9 64 19 3.4 1.6 0.5 TR7 2005 9 62 19 3.3 0.7 -0.3 TR8 2005 9 65 18 3.6 1.0 0 TR9 2005 9 65 18 3.6 3.0 1.1 TR10 2005 9 67 18 3.8 3.6 1.3 TR11 2005 9 65 19 3.4 1.1 0.1 TR12 2005 9 67 23 3.0 0.8 -0.2 TR13 2005 9 66 20 3.3 0.8 -0.2 TR14 2005 9 64 16 4.1 4.0 1.4 TR15 2005 9 66 15 4.3 4.6 1.5 TR16 2005 9 72 25 2.8 1.0 0 TR17 2005 9 65 16 4.1 3.8 1.3 TR18 2005 9 65 16 4.0 5.4 1.7

Water-quality data were obtained from the Lakes of total suspended solids according to procedures established by Missouri Volunteer Program (LMVP). This program consists LMVP (Obrecht, 1992). Samples are collected once every 3 of citizen volunteers who are trained to collect and process weeks between April and September for 18 sites (fig. 2) on water samples from Missouri lakes. Volunteers monitor Table Rock Lake. Repeat sampling is conducted by the same surface-water temperature, water clarity (secchi disk trans- volunteer at each site with site occupation mostly depen- parency), chlorophyll, total phosphorus, total nitrogen, and dent on aligning with visible landmarks. This method might Methods 5

7.0

6.0

5.0

4.0

3.0

2.0 MEASURED SECCHI DEPTH, IN METERS

1.0

0 0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 PREDICTED SECCHI DEPTH, IN METERS

Figure 3. Relation between predicted and measured values of secchi depth for Table Rock Lake. account for some error in consistently occupying the same with potential vegetation, shoreline, or bottom effects. Mean sample location, but most volunteers are local residents and brightness values for each TM band were collected for each have an intricate knowledge of their sampling locations. sample site AOI (table 1). Processing of the satellite data was done using ERDAS Regression analysis was performed using SPSS 14.0 for Imagine software and modified from the procedures described Windows (SPSS Inc., 2005). The following regression equa- by Olmanson and others (2001). The first step is to produce tion developed by Kloiber and others (2002) was used in the a water only image so that areas outside of the water body analysis of the two scene dates: are eliminated from the analysis area. This is accomplished by performing an unsupervised classification on the Landsat image to differentiate water pixels from all others. The scene ln(SDT) = a(TM1/TM3) + bTM1 + c (1) can then be recoded to create a binary image representing water and terrestrial areas. The water only image is further processed using an unsupervised classification to separate where deep, open water areas from areas of the lake affected by ln(SDT) is the natural log of secchi disk transparency vegetation, shoreline, or bottom effects. These areas should be depth, in meters, computed from the avoided when choosing sample points in the satellite scene. LMVP water quality data; Sample sites for satellite signature acquisition were a, b, and c are the coefficients derived from the selected to coincide with the LMVP water-quality sites. A 3 regression analysis; and by 3 (9 pixel) area of interest (AOI) was created for each of TM1 and TM3 are bands 1 [0.45-0.52 micrometers (µm)] and the sample sites. The AOI size was determined based on a 3 (0.63-0.69 µm), respectively, from the previous study by Kloiber and others (2002) that showed cor- Landsat 5 TM. relation strength improved with an increase in AOI size from 1 to 9 pixels, but was marginal when the AOI was increased beyond 25 pixels. AOIs were centered over the LMVP sample site location and visual inspection performed to avoid areas 6 Table Rock Lake Water-Clarity Assessment Using Landsat Thematic Mapper Satellite Data

4 NUMBER OF OBSERVATIONS NUMBER OF OBSERVATIONS 3

0 -4 -3 -2 -1 0 1 2 3 4

DAYS FROM SCENE ACQUISITION

Figure 4. Distribution of field samples with respect to number of days within satellite overpass. All field samples were collected within 4 days with the largest percentage of samples collected 3 days from the time of satellite data collection. Results percent), 3 days (44 percent), and 4 days (25 percent). This distribution is inherent in the volunteer sample program as no effort is made to collect all samples on the same day. The regression analysis based on equation (1) produced Quality and reliability of volunteer collected data have the following linear model: been evaluated by Obrecht and others (1998). They found no statistical differences in data collected by volunteers and data ln(SDT) = -10.061(TM1/TM3) + 1.788(TM1) + 0.064 collected by University of Missouri personnel for total phos- (2) phorus, total nitrogen, chlorophyll, or secchi depth. with a reliability (R2) of 0.76 and standard error of the esti- Results of the regression analysis were used to create mate of 0.314. These results are consistent with similar studies models using Imagine Spatial Modeler to output images show- using the TM1 and TM3 band relation to secchi disk transpar- ing the ln(SDT) and trophic state index (TSI) derived from ency (Kloiber and others, 2002; Lathrop, 1992). Figure 3 is ln(SDT) for the entire lake. The relation between TSI and sec- a plot of the relation between predicted and measured secchi chi disk depth is defined as: depth for sample locations on Table Rock Lake. The plot shows a strong linear relation (r=0.83) between measured TSI = 60-14.41 ln secchi disk (3) secchi depth and secchi depth predicted from mean Landsat reflectance values. where Field data collected on the same day as the satellite over- TSI is the Carlson trophic state index, and pass result in the best regression results. Data collected from 1 ln secchi disk is the natural logarithm of the depth to 7 days off the overpass date result in a decreasing strength measurement, in meters. of correlation (Kloiber and others, 2002; Chipman and others,

2004). All of the field data for this study were collected within Output images are pixel-level representations of transparency 4 days of the satellite overpass (fig. 4) with a distribution as or trophic state that can be used to examine differences in follows: same day (0 percent), 1 day (3 percent), 2 days (28 Results 7

EXPLANATION

Universal Transverse Mercator projection, zone 15 0 5 10 MILES Horizontal coordinate information is referenced to the North American Datum of 1983 (NAD 83) 0 5 10 KILOMETERS

Figure 5. Model predicted trophic state for Table Rock Lake – August 2003. Water quality is degraded in the main tributaries compared to the main channel of the reservoir.

EXPLANATION

Universal Transverse Mercator projection, zone 15 0 5 10 MILES Horizontal coordinate information is referenced to the North American Datum of 1983 (NAD 83) 0 5 10 KILOMETERS

Figure 6. Temporal comparison of model results for the James River arm of Table Rock Lake. Water clarity has decreased in 2005. 8 Table Rock Lake Water-Clarity Assessment Using Landsat Thematic Mapper Satellite Data water quality throughout the reservoir (fig. 5). Differences in Chipman, J.W., Lillesand, T.M., Schmaltz, J.E., Leale, J.E., water quality are readily apparent between the main channel and Nordheim, M.J., 2004, Mapping lake water clarity with and major tributaries with darker green and yellow pixels in Landsat images in Wisconsin, U.S.A.: Canadian Journal of the tributaries representing decreased water clarity. Temporal Remote Sensing, v. 30, p. 1–7. comparisons can also be made to assess changes in water clarity over time. Figure 6 shows the difference in water clarity Fuller, L.M., Aichele, S.S., and Minnerick, R.J., 2004, Predict- for the James River arm of Table Rock Lake between the 2003 ing water quality by relating secchi-disk transparency and and 2005 scene dates. The increase in the TSI value can be chlorophyll a measurments to satellite imagery for Michi- seen extending much farther south in the James River arm in gan inland lakes, August 2002: U.S. Geological Survey the 2005 image. Scientific Investigations Report 2004–5086, 25 p. Jones, J.R., and Perkins, Bruce, 1999, Table Rock Lake: An evaluation of factors regulating its trophic state, Final Summary Report to Missouri Department of Natural Resources and U.S. Environmental Protection Agency. University of Mis- Remote sensing techniques can be effective in monitoring souri, 82 p. water quality in southwestern Missouri. Mean Landsat bright- Kloiber, S.M., Brezonik, P.L., Olmanson, L.G., and Bauer, ness values can be used with field data collected by citizen M.E., 2002, A procedure for regional lake water clarity volunteers to produce reliable predictive models of water assessment using Landsat multispectral data: Remote Sens- clarity and trophic state based on measurements of the secchi ing of Environment, v. 82, p. 38–47. depth transparency. Predictive models can be used to supple- ment standard water-quality analyses during periods when Krizanich, G.W., 1986, Landsat trophic state assessment of field data are not being collected by volunteers. Temporal Fellows Lake: Missouri State University, Masters thesis, analyses are possible through the exploitation of the Landsat 64 p. archive. Pixel-level representations of water clarity and trophic state provide a means of visually interpreting the status of Lathrop, R.G., 1992, Landsat thematic mapper monitoring of water quality. turbid inland water quality: Photogrammetric Engineering and Remote Sensing, v. 58, p. 465–470. Lathrop, R.G., and Lillesand, T.M., 1986, Utility of thematic Acknowledgments mapper data to assess water quality in southern Green Bay and west-central Lake Michigan: Photogrammetric Engi- Field data used in this study was provided by the Lakes neering and Remote Sensing, v. 52, p. 671–680. of Missouri Volunteer Program. Special thanks go to Tony Missouri Department of Natural Resources, 2005, Table Rock Thorpe for his help in acquiring the original data and for Lake total maximum daily load information sheet: Missouri discussions on procedures and interpretations of the volunteer Department of Natural Resources Water Protection Pro- data. gram, 4 p. Obrecht, D.V., 1992, Lakes of Missouri Volunteer Program References Cited Manual: Lakes of Missouri Volunteer Program, Universtiy of Missouri, 17 p.

Allee, R.J., and Johnson, J.E., 1999, Use of satellite imagery Obrecht, D.V., Milanick, Margaret, Perkins, B.D., Ready, to estimate surface chlorophyll a and secchi disk depth of Diana, and Jones, J.R., 1998, Evaluation of data generated Bull Shoals Reservoir, Arkansas, USA: International Jour- from lake samples collected by volunteers: Journal of Lake nal of Remote Sensing, v. 20, no. 6, p. 1057–1072. and Reservoir Management, v. 14, no. 1, p. 21–27. Boland, D.H.P., 1976, Trophic classification of lakes using Olmanson, L.G., Kloiber, S.M., Bauer, M.E., and Brezonik, Landsat-1 (ERTS-1) multispectral scanner data: U.S. Envi- P.L., 2001, Image processing protocol for regional assess- ronmental Protection Agency, Corvallis, Oreg., EPA-600/3- ments of lake water quality: Water Resources Center and 76-03, 245 p. Remote Sensing Lab, University of Minnesota. Carlson, R.E.,1977, A trophic state index for lakes: Limnology Ritchie, J.C., Zimba, P.V., and Everitt, J.H., 2003, Remote and Oceanography, v. 22, no. 2, p. 361–369. sensing techniques to assess water quality: Photogrammetric Engineering and Remote Sensing, v. 69, no. 6, p. 695–704.

SPSS for Windows, Release 14.0.1, 2005, Chicago: SPSS Inc. References Cited 9

Thorpe, A.P., Obrecht, D.V., and Jones, J.R., 2005, Lakes of Missouri Volunteer Program 2005 data report, Department of Fisheries and Wildlife Sciences, University of Missouri. Wetzel, R.G., 2001, Limnology: lake and river ecosystems (3d ed.): San Diego, CA, Academic Press, 1006 p. Wezernak, C.T., Tanis, F.J., and Bajza, C.A., 1976, Trophic state analysis of inland lakes: Remote Sensing of Environ- ment, v. 5, p. 147–165. Publishing support provided by: Rolla Publishing Service Center

For more information concerning this publication, contact: Director U.S. Geological Survey Mid-Continent Geographic Science Center 1400 Independence Road, Mail Stop 236 Rolla, MO 65401 (573) 308–3593 Or visit the Mid-Continent Geographic Science Center website at: http://mcgsc.usgs.gov/

Krizanich and Finn—Table Rock Lake Water-Clarity Assessment Using Landsat Thematic Mapper Satellite Data—Scientific Investigations Report 2009–5162